I have managed to translate my code to CUDA and the speedup is 5x more which is great for a piece of code that uses global memory.
Now my only problem is the fact that since I can’t use static variables within device functions, it has rendered the Random Number Generators useless. The formulas were obtained from ‘Numerical Recipes 2nd Ed’.
They are: gammaln(), poidev() and ran1().
Can anyone suggest an alternative? It’d be great if they are seedless.